ae eesd in Mathematical Physics Dariusz Chruscinski Andrzej Jamiotkowski Geometric Phases in Classical and Quantum Mechanics BirkhauserProgress in Mathematical Physics Dariusz Chruscitiski ‘Volume 36 : 7 : Andrzej Jamiotkowski Editors-in-Chief Anne Boutet de Monvel, Université Paris VII Denis Diderot Gerald Kaiser, The Virginia Center for Signals and Waves Editorial Board . Sit M. Bery, University of Bristol Geometric Phases C. Berenstein, University of Maryland, College Park 7 P. Blanchard, Universitat Bielefeld ] AS. Fokas, Imperial College of Science, Technology and Medicine | in Classical and D. Sternheimer, Université de Bourgogne, Dijon | é Cotacy, University of California, Davis Quantum Mechanics Birkhauser Boston + Basel « BerlinDarius2 Chrasciiski Nicholas Copernicus University Andrzej Jamiokkowski Nicholas Copernicus University ‘Torun 87-100 “Torun 87-100 Poland Poland Libeary of Congress Cataloging in-Publication Data (Chrseist, Dares. ‘Geometi paces in cls and quantum mechanics / Dariusz Chris, Andre Jamin ‘p.m. =(Progresin mathemati payies ¥.36) Includes bibliographical references and index. ISBN 0'8176-1282-X alk. pape 1. Geometric quanti phases. 2, Quntum theory 3, Mathematical physic. Samiolkowsk, Andre) 1846-1 Tide. IL Seis (9020.7.644C87 2004 S30.12-de22 ooro46aT8 : cr [AMS Subject Classifications 29605, 29670, 4708, 37910, SIS, 3755, 3301 SAAT, SID, 53820, S3B21, $3835, $3850, $3005, SICIT, $3C10, S3C15, 83C22, 33C29,$3C30, CSS, S3C30, 580s, 53020, 53022, 53205, ROS, S5R 10, 55R2S,SSROI, SSADS,SBA10, SBA2, $8505, 70ELS, “04m, 70806, TORT, TORS, 7HA25, 98A35,S1PSB,S1QOS, BLOT ISBN 0-8176-4282-% Printed o aid tree pope. (©2004 Birkhiaser Boston Birkhauser IB ‘Allright reserved This work nay not be raatedor copied in bole o in par witout he writen permission of the publisher Bckhlanee Boston Springer Vera New Yor, LLC. 178 Fit ‘Avenue, New York, NY 10010, USA) except for re exces in connection with eviews or scholarly analysis, Use in connection wit ny form of infomation storage and retical, electoie ‘Rapation, computer software, or by similar or dsinlar mthodology now known or hereafter’ ‘eveoped is feridden, The se this publication of trade nates, aemars servicemarks 2nd Similar ems, even if hey are no ered ssh, nto Be ake a a expression of opinion a tower ono they are subject propery res Print inthe United States of America. (TXQMV) pe7esaaai ‘SPIN 10863573 Birkioscr ea pa of Springer Sciences Business Media ew birasercom Contents Preface 1. Mathematical Background 12 Manifolds, forms and all that LLL Basic notions 1.12 Differential forms . 1.13 Integration of forms 1.14 DeRhamcohomology 115 Lie derivative . Groups. Lie algebras and actions. 1.2.1 Basic definitions 1.2.2 Actions of Lie groups 1.2.3 Homogeneous spaces 1.24 Lie algebras and differential forms Bundles and connections . 13.1 Fibre bundle 1.3.2 Examples of fibre bundles 1.3.3 Connections — general theory 13.4 Connection in a principal bundle 1.3.5 Gauge transformations — paving the way to physics 1.3.6 Characteristic classes ‘Topology, bundles and physics 1.4.1 Elements from homotopy theoryvi Contents 1.42 Monopole bundle 143 Tnstanton bundle 1.4.4 Hopf fibration $3 —> s? 145 Hopf fibration 57 —> 5* Problems 2 Adiabatic Phases in Quantum Mechanics 2.1. Adiabatic evolution in quantum mechanics 2.1.1 Adiabatic approach of Born and Fock 2.1.2 Adiabatic theorem 241.3 Adiabaticity and geometry 2.2 Berry's phase 2.2.1 Phase in quantum mechanics 22.2 Standard derivation . 22.3 How to measure the Berry phase . 2.2.4 Berry-Simon connection 225 Examples 2.2.6 Quantum geometric tensor 22.7 Quantal phase and geometry — a simple illustration , 2.3 Non-abelian Wilezek~Zee phase 23.1 Standard derivation... +... 23.2. Fibre bundle approach 2.3.3 Non-abelian phase in quadrupole resonance Problems sees : 3. Adiabatic Phases in Classical Mechanics 3.1 Hamiltonian systems . 3.1.1 What we mean by a phase in classical mechanics 3.1.2 Symplectic geometry and Hamiltonian dynamics 3.1.3. Integrable systems. 3.2 Adisbatic phase of Hannay 3.2.1 Averaging principle 3.2.2 Adiabatic invariants 3.23 Hannay’s angles 3.24 Berry's phase versus Hannay’s angle . 3.3. Classical geometric phases — examples 33.1 “Classical spin” 33.2 Classical oscillator 3.3.3. Rotated rotator (Hannay’s hoop) 3.3.4 Motion in non-inertial frames 3.35 Guiding center motion 3.3.6 _ Rigid bodies and the geometric phase Problems : . 7 a 68 67 B B B 16 B » 83 86 96 100 100 101 10s 108 mL mL 1 1B 7 121 121 123 128 130 132 132 136 138 142 146 149 154 Contents 4 Geometric Approach to Classical Phases 4.1 Hamiltonian systems with symmetries 4.1.1 Hamiltonian actions and momentum maps 4.1.2 Reduced phase space 4.2 Geometric approach to adiabatic phases 4.2.1 Families of hamiltonian actions 4.2.2 Hannay’s angles and the Hannay-Berry connection 43 Reduction, reconstruction and phases 4.3.1 Reconstruction of dynamics 43.2 Rigid bodies Problems Geometry of Quantum Evolution 5.1 Geometrical formulation of quantum mechanies 5.1 Hilbert space as a Killer manifold 5.1.2 The quantum phase space 5.1.3 Example: geometry of CP" 5.1.4 Symplectic structure and quantum dynamics 5.15 Metric structure and uncertainty relation 5.2. Aharonov-Anandan phase 5.2.1 Standard derivation 5.2.2 Example: spin-half in a magnetic field 5.2.3 Fibre bundle approach 52.4 Geometry of CP" and Aharonov-Anandan phase 5.3 Quantum measurement and Pancharatnam phase . 5.3.1 Geodesies in quantum phase space 5.3.2 Distance between pure quantum states 53.3 Measurement process 53.4 Pancharatnam phase 5.4 Geometric phase for mixed states . 5.41 Mixed states in quantum mechanics 5.4.2 Uhlmann’s non-abelian geometric factor 543. Distance between mixed states 5.44 How to measure geometric phase Problems, Geometric Phases in Action 6.1 Optical manifestation of geometric phases 6.1.1 Spins and helicities 6.1.2 Chiao~Tomita-Wu phase 6.13 Rytov’s law and Fermi-Walker teansport 6.14 Pancharatnam phase 62 Quantum mechanics as a gauge theory 6.2.1 Classical particles in gauge theory 187 157 157 160 163 163 165 168 168 13 178 179 179 179 134 187 190 193 196 190 200 202 204 + 205 208 21 213 218 218 221 27 232 236 239 239 239 242 243, 247 249 250Contents 6.2.2 U(1)-invariance and the Aharonov-Bohm effect 62.3 SU@)-invariance and the Aharonov-Casher effect 6.3. Phases in molecular physics 63.1 Degeneracies 6.3.2 Time-reversal invariant fermionic system 63.3 Bom-Oppenheimer approximation 634 Diatomic molecule 6.3.5 Quantum geometric forces 6.4 ‘The quantum Hall effect 64.1 Preliminaries . 642 Quantum dynamics in a magnetic field 6.4.3 Fibre bundle approach to the QHE 6.5. Spin, statistics and the geometric phase 65.1 ‘The transported spin basis, 6.5.2 Schwinger representation 65.3 Exchange rotation : 65.4 Pauli signas a topological phase 66 Entanglement and holonomic quantum computation . 66.1 Composite systems and entangled states 6.6.2 Qubits and bundles 6.63 Quantum computer — an overview 6.64 Holonomic quantum computation Problems Classical Matrix Lie Groups and Algebras B-Quaternions Bibliography Index 251 254 258. 258 261 266 + 269 21 214 214 21 279 281 282 285 288 292 293, 293 295 297 300 303 3a Preface Twenty years ago Michael Berry (Berry 1984) demonstrated that the standard descrip tion of adiabatic processes in quantum mechanics is incomplete. Berry noticed that if the Hamiltonian ofthe system depends on a numberof parameters which vary adiabat- ically during the evolution, then a cyclic variation of the parameters is accompanied by a change of wave function by an additional phase factor, which hitherto had boen ‘completely ignored. This additional factor, known today asthe Berry phase, possesses a remarkable geometrical property —it depends only upon the geometric structure of the space of parameters and does not depend on the duration of the evolution. ‘Almost simultaneously, a similar phenomenon was observed by J. Hannay (Hannay 1985) within the framework of classical mechanies,' The classical counterpart of the phase of the wave function isthe phase of quasi-periodic motion — the so-called an- gle variable in the action-angle representation of integrable systems. These classical geometric phases are called Hannay's angles. It was soon realized that Bery’s remarkable observation finds a surprisingly broad spectrum of applications. An analog of Berry's geometric phase is manifested in many apparently unrelated phenomena like, for example, the Foucault pendulum, the passage of photons through optical fibres, the spectra of molecules, the quantum Hall effect, and anomalies in quantum field theory. Why is this so? Tt turns out that they have a universal mathematical description. B. Simon (Simon 1983)? was the first who ob- served that Berry's phase may be interpreted as a purely geometric object, namely as a "Interestingly. many imporant discoveries connected with the notion of gometric phase were made in Bristol For the Bristol Anhonomy Calendar see Berry 1991, and the review Bery 19908, 2 Acually, de 1o the referee's delay Simon's paper appeared in 1983 and Bey 1984x Preface holonomy in a certain fibre bundle. This way an elegant and fairly sophisticated mathe- matical theory of bundles and connections enters elementary quantum mechanics. The bundle constructed by Simon, called by physicists the spectral bundle, is uniquely de- termined by the spectral properties ofthe system's Hamiltonian, Simon showed that the adiabatic evolution considered by Berry defines the so-called parallel transport of ‘vector from the system's Hilbert space along a curve in the parameter space. Now, & vector transported along a closed curve does notin general return to its original form ‘but acquires a phase factor predicted by Berry. This geometrical phenomenon is well known in classical geometry: a vector parallel transported along a closed curve on a two-dimensional sphere does in general change its orientation by an angle of rotation ‘equal to the solid angle subtended by the curve. AS we shall see, this 19th century ‘observation finds new. interesting applications in various branches of physics. ‘Actually, the theory of fibre bundles was successfully applied in the seventies (ofthe last century) in the mathematical formulation of gauge theories — electrodynamics and its generalization known as Yang-Mills theory. Therefore, itis not surprising that there are many analogies between gauge theories and geometric phases. For example, such classical topics as magnetic poles and instantons find new and fresh illustration when dealing with geometric phases. ‘There are several reviews that focus on various aspects of geometric phases. Berry's articles Quantum Adiabatic Holonomy (Berry 1989a) and The quantum phase, five ‘ears after (Berry 1989) constitute a beautiful overview and summarize the firs, most ‘exciting period in the development of the subject. The articles Topological phases in quantum mechanics and polarication optics (Vinitskii etal. 1990) and Polarization of light and topological phases (Bhandari 1997) presents the application of geometric phases in optics. Phases in molecular physics are reviewed in The geometric phase in molecular systems (Mead 1992) and The Geometric Phase in Quantum Mechan ‘cs (Bohm 1993b), Actually, the third edition of Bohm’s Quantum Mechanics (Bohm 1993a) contains an introduction tothe subject together with examples from molecular Physics. Moreover, there are two collections of papers: Topological Phases in Quan- ‘um Theory, edited by Markowski and Vinitski, and Geometric Phases in Physics with excellent commentaries by Shapere and Wilczek. Finally, there are the reviews Berry hase (Zwanzinger etal 1990) and Geometric phases in physics (Anandan etal. 1997), ‘hich provide very useful guides tothe literature on geometric phases. Although in- terest in the geometric phase dates to the mid 1980s, the beginnings of this problem came much earlier. Exciting historical reviews may be found in Anticipations of the ‘geometric phase (Berry 1990b) and the appendix of Topological phases in quantum mechanics and polarization optics (Vinitskii et al. 1990). Actually, when this book was completed there appeared a monograph, The Geometric Phase in Quantum Sys- ‘ems by Bohm et al. (Springer 2003), with a thorough introduction and applications from molecular and condensed matter physics. Why write yet another book on geometric phases? What distinguishes this book fom other texts is that it covers both quantal and classical geometric phases from a unified, geometric point of view and at a rather sophisticated level. Moreover, it pro- ‘ides insights into the relationships between quantal and classical phases which have Preface xi not been emphasized previously atthe textbook level. This book is addressed to grad- uate students in mathematical and theoretical physics, as well as theoretical physicists ‘and applied mathematicians. We hope that it helps the reader to enter the exciting world of geometric phases in classical and quantum mechanics and to feel their elegant, co- hherent mathematical description. It can certainly be used as a supplementary textbook in a course on differential geometry for physicists, as well. ‘To achieve our goal we start with the mathematical background in Chapter 1 and plunge the reader into the arena of differential geometry. This introductory chapter also includes basic facts from Lie groups and algebras and concentrates on fibre bun- dies and connections, which serve as the main tool to study geometric phases. The reader will find a detailed exposition of the celebrated Hopf fibrations and several physical illustrations as magnetic poles and instantons in Yang-Mills theory. Having, at hand the mathematical tools, we startin Chapter 2 to investigate the physics of ge- ‘ometric phases. The reader learns about adiabatic theorem in quantum mechanics, the ‘Quantal adiabatic phases of Berry, and their non-abelian generalization due to Wilezek and Zee. The presentation of the physical side of the problem is simple and requires only basic notions from quantum mechanics; it should be accessible for mathemati cians interested in theoretical physics. The mathematical side uses the previously in- troduced notions of fibre bundles and connections, and stresses the geometric aspects of adiabatic evolution, Chapter 3 deals with adiabatic geometric phases in classical mechanics. Here we in- lroduce basic facts from symplectic geometry, Hamiltonian mechanics and integrable systems. It is shown how the classical adiabatic theorem leads to classical geometric phases — Hannay's angles. This chapter also includes many examples of classical sys- tems displaying geometric phases. In Chapter 4 we present the geometric approach to Classical phases using the mathematical language of bundles and connections. It in- {roduces elegant geometric constructions: momentum maps, the celebrated Marsden— ‘Weinstein reduction procedure and finally the Hannay Berry connection. This chapter is illustrated by the dynamics of the rigid body — a system where the analog of the geometric phase was already observed in the 19th century. ‘Chapter 5 describes the natural geometric structure of quantum evolution, The stan dard approach to nonrelativistic quantum mechanics is based on a complex Hilbert space. However, as is well known, the Hilbert space is not an appropriate phase space for the quantum system. Any two unit vectors differing by a phase factor define the ‘same physical state and hence they are physically equivalent. Therefore, the rue quan- ttum space of states is a projective Hilbert space — the space of equivalence classes (or the space of rays). A projective Hilbert space is endowed with two geometric struc- tures — a Riemannian metric and a symplectic form. Hence, the geometric structure ‘of quantum evolution is much richer than its classical counterpart, We show that sym- pleetic structure is responsible for the so-called Aharonov-Anandan geometric phase ‘whereas metric structure is closely related to the beautiful notion of the Pancharat- nam phase. Finally, we present the geometric framework for the quantum evolution of, ‘mixed states and the corresponding nonabelian Uhlmann geometric phase. The recent Proposal of interferometric measurement of the geometric phase for mixed states isxii Preface also included. The geometric framework applied to quantum mechanies enables one to get more insight into the structure ofthe quantum space of states and deserves to be more widely known, We stress that this topic has not been so extensively discussed in the literature on geometric phases. Recently, it received considerable attention due t0 the rapid development inthe field of quantum information theory. Finally, Chapter 6 shows the geometric phases “in action.” It includes several stan- dard examples such as the appearance of geometric phases in optics and molecu- lar physics. We present coherent derivations of the Aharonov-Bohm and Aharonov— Casher effects using the underlying symmetries of nonrelativistic quantum mechanics. We show how the geometry of fibre bundles enters the highly nontrivial physics of {quantum Hall effects and show how topology explains the quantization of Hall conduc- tance, Moreover, we review Berry and Robbins’ (Berry and Robbins 1997) approach to the spin-statistics theorem, We close this chapter witha discussion of the recent sur- prising application of geometric phases to quantum computation — geometric phases are used to model quantum gates in a quantum computer. This example shows that the subject of geometric phases is still alive and perhaps one can see just another manifes- tation of geometric phases in the near future. Acknowledgements. The authors thank Professor 16et Szuy, the dean ofthe Faculty Of Physics, Astronomy and informatics a Nicolaus Copemics Univesity in Torus for hs ines n this project and his Lind support We thank Ava Kostant and the staff t Birkhioser Boston fr their advice and help. Special thanks are duc to David Winters for his strate copyediting and to Elizabeth Loew for her excellent production of ths took D.C. woul like to thank Jerzy Kijowski and Gerry Kaiser for their encouragement during tis project. 1 thank professor Iwo Bialynickt-Birula for his remarks on the history of geometric phase in optics. am grateful to Hartmann Rémer for many in teresting discussions and his kind hospitality in Freiburg, where part of the work was Gane. Special thanks are due wo Jacek Jurkowski who carefully read the maniscip and pointed out many errors. For nancial support I thank the DFG-Graduiertenkolleg Nichlineare Diferenalgleichungen in Freiburg and the Nicolaus Copericis Univer sity in Torus Grant nr M/72001). Most ofall, thank my wife Alia for her love, Patience and understanding. Dariusz Chrusciiski Andrzej Jamiotkowski Toruti, October 2003, Geometric Phases in Classical and Quantum Mechanics1 Mathematical Background ‘The mathematical background required for the study of geomet phases in classi- cal and quantum mechanics is rather extensive, The aim of this introductory chap- ter is to provide 2 background of some basic notions of classical differential geom- cety and topology. Classical differential geometry is now a well established tool in ‘modern theoretical physics. Many classical theories like mechanics, electrodynamics, Einstein's General Relativity or Yang-Mills gauge theories are well known examples where the geometrical methods enter in the natural and very effective way. As we shall see throughout this book, also quantum physics shows its intricate beauty when ‘one applies an appropriate geometric framework. All this proves Wigner's celebrated statement about the “unreasonable effectiveness” of mathematics in natural sciences, 1.1 Manifolds, forms and all that 1.1.1 Basic notions ‘The concept of a manifold generalizes the concept of a smooth surface or a curve in ?. Manifolds occur in all areas of physics. Whenever one speaks about spaces, like ‘space of states, configuration space, parameter space or a physical space-time, one usually deals with manifolds A topological space M is called an n-dimensional topological manifold if it looks locally like a Buclidean space R", or, more precisely if there exists a family of open subsets (Uj, € 1) of M such that 1. itcovers M, ie Uses Ui = Ms21, Mathematical Background 2. for each i € 1 there is a homeomorphism (continuous, invertible map) Us > eW) CR" ‘A pair (Ui, %) is called a chart or more often a local coordinate system. Since U; —> RY, we may represent gas follows GE = ly Rye AM)» where xf, : Uy > R, fork = 1,2,..- m. The maps (4p. 42). ..< 2%) are called local coordinates on the patch U;. Moreover, any two chars on Mf have to be compat- ible, thats, the overlap map PH = 9) 08) + CMULDU)) — @UL NU), an defines homomorphism for any i,j € J, ef. Fig. 1.1. An overlap map yi describes hhow the different chart are glued together. One calls a collection of compatible charts covering M an atlas. If the maps g;i are C™ diffeomorphisms, ie., they are N times differentiable, then M is called a CN manifold. In particular, C' and C°° manifolds ate called differentiable and smooth manifolds, respectively. Throughout this book all manifolds are assumed to be smooth. ‘The most trivial example ofa diferentiable manifold is an open subset U of R". An aulas consists just of one chart (U, dy), where idy stands for the identity map on U. However, usually one needs more than one chat o cover M. An important example is provided by Example 1.1.1 (n-dimensional sphere 5") 5” is defined as the following subset in Rel stm [at, vey ent | Son 1} fat ailed Figure 1.1: Overlapping chars on a manifold 1.1, Manifolds, forms and all that 3 Define two patches Uy and Us covering S" S*~((0,0,....0,-D), S*={0,0,...,0,1)) Us = fe! “al In analogy with R? we may call the points {(0,0,... ,0,—1)) and {(0,0,... ,0, D} th sou an nor pols ope Non et we en Wo apy Uy me and gs Us > Re by deren precion ene ea plane x"*! = 0 from the south and north poles, respectively. That is, owt xt) Gh) pet woCh oe A) me a) (ne easily nds hat the overlap map ys, given by - (&«») Oh) defines a smooth map on R — {0}. This proves that the two charts (Uy. gy) and Ws.) provide an atlas on S" 5 ense', ‘An atlas on a manifold M enables us to develop a differential calculus, A function fon M, is,,amap f: M —> R, is differentiable it Foye! : (Ui) — RY (2) is differentiable for each chart (Ug). We shall denote the space of smooth functions fon M by C™°(M): Consider now a chart (U.g'= (r,s. .x")) on M and letx € U. Define a differential operator Vs atx by vin=dvi Ley, as ir for any differentiable function f on U. This construction shows that Vis uniquely determined by n numbers (V,..., Vf) and hence that differential operators atx span an n-dimensional linear space called the tangent space T,.M atthe point x. Elements from T,M are called tangent vectors (attached) at the point x. Ths, there isa one-o- ‘one correspondence between differential operators and tangent vectors. Clearly the st of differentials : 21} as) Bartle at le41, Mathematical Background efines a basis in T;M — the so-called coordinate bass. Then numbers (V2... Vf) fare components of a tangent vector V; with respect f0 a coordinate basis. From its definition itis clear that Va(f) i the directional derivative of a function fata point x in the direction of a vector Vz. An assignment Moxy eTM, as) is called a vector field on M. We shall denote the space of vector fields on M by X(M). Denote by 77M the algebraic dual of 7M, called a cotangent space atx. Physicists often call elements from 7; M covectors (dual vectors). Having a coordinate basis (1.4) in T,M let us introduce the corresponding dual basis in 77M: fasta} ao such that, we (2) =a, an ‘The above constructions of tangent and cotangent spaces enable one to introduce an arbitrary tensor field on M. We shall call a smooth map k 1 Max — Ta) € TM =TM@..@TMOTM@..OTM as) a tensor field of type (&, 1) Clearly, a vector fed i a tensor field of type (1,0). Any tensor field is uniquely defined by its components. In particular, using a coordinate basis in 7, M and TM we have ‘ioe 8 2 gash A T = Tick 5 @...@ 5, deh @...@dr!, a9 Where we use the Einstein summation convention, 6 Let us consider two manifolds M and N together with a smooth map e:M—N ‘The derivative Tyw (one calls it also a tangent map) of g ata point x & M is a linear map Teg: TeM —> Tyo), such that IeW)IA) = ue. 10) 1.1. Manifolds, forms and all that S forany f € C™(N) and vy € TyM. Representing g in local coordinates (x1,.... x" PO), ¢ the tangent map Ty¢ is represented by the m x m matrix ag! we. i (er, Using a map y ; M —> V one may transport tensor fields between M and N. Let @ be a tensor field of type (0, k) on WV. A pull-back g*w of « is a (0,k)-tensor field on M defined by W)C... VE) = Wye (Te(01),--. Te l04)) + «uy for any x € M and v1,... , o4 € TeM. Conversely, a tensor field U of type (J, 0) may bbe pushed forward from M to N, giving rise to an (I, 0)-tensor field y4U GU gcaylory--. 04) = Urlefen,... gran), (1.12) where a... .ay € TiN. This easy to show that if e : M —> N is a diffeomor- phism, then w= a.) Example 1.1.2 Let V be a (1, 0)-tensor on M, that is, a vector field V = V/A/ax/ ‘Then the pushed-forward vector field y.V is given by a OV = VE ‘where (3/2y/} denotes the coordinate basis on N and a VO) = ew via Similarly, if is a (0, 1)-tensor on N,, .e.,a covector field a = a; dy', then the pulled- back covector field y*a' on M is given by os ea); dx! , with @'o,/@) = Serato) ‘This way, one recovers the well-known transformation rules for vectors and covectors °6 1. Mathematical Background 1.1.2. Differential forms Now we are going to introduce an important class of tensor fields on a differential ‘manifold that play a prominent role in physical applications. Definition 1.1.1 A skew-symmetric tensor of ype (0, k) is called a differential form of onder k (or simply a k-form) Denote by A¥(M) the space of k-forms on M. Evidently, A(M) = (0} for k > n. Therefore, the space of differential forms on Mf, denoted by A(M), splits into the following direct sum: Aah =@atan, with A°%CM) = C™(M). The space AM) is equipped with two basic operations: wedge product and exterior derivative, Definition 1.1.2 A wedge product - (called also the exterior or Grassmann product): At ARM) x ATM) > akan), ag) is defined by ato! ka! Where A isan alternation operator which selects the skew-symmetric part of the (0, + D-tensor a @ B. apm A@ep), ats) For example, if « and f are one-forms, then (vB (v2) ~ a(e2)B(01) , 16) for any vectors vy and vp. In terms of local coordinates (x, .. ,") any k-forma has the following component representation: A BYU, ») = eds 1X cconds ay One easily shows that (OW Bigndeas = Mii Bayrategd + Where the square bracket stands for anti-symmetrization. Example 1.1.3 Ifa = a) ds! and § = By ds/, then + a8) AB =eifj dx! nds! = MasBy —ajfi)ds! nde! and hence (@AB)j = eB; ~ 0/6 in agreement with formula (1.18) ° 1.1. Manifolds, forms and all hat 7 ‘The properties of the wedge product are summarized in the following. Proposition 1.1.1 The wedge product satisfies 1 @AB)AY =aNBAY), 2. @AB=(-1)B Aa, wherea € AK(M) and B & A'(M). Since there isa natural pairing between vectors and forms it is convenient to introduce the so-called interior product, which is a contraction of a vector field v ¢ X(M) and a form a fy: AFM) — atl), that is, Go Mister = Werte 19) Moreover, we declare that iy f =O for f € A°(M), ‘The next operation we are going to introduce enables one to differentiate k-forms. Definition 1.1.3 The exterior derivative a: ka) — aon) is defined as follows: Me dd nbn onda, am for any k-form a represented by the formula (1.17). Example 1.1.4 Consider a function f on a differential manifold M. Since a function is a zero-form one defines its exterior derivative df to be the following one-form: a apm Las! Note that, aap) = oF de! nde! =0, because the partial derivatives #4) f are symmetric in (ij) whereas dx! Ads" is anti- symmetric. In particular, if M = R® and (x!,... ,x") are cartesian coordinates, then df reproduces the components of grad f. ° Example 1.1.5 (Differential forms in R*) Let us choose cartesian coortinates (e!, x7.23) in R? and consider a one-form a and a two-form . Clearly, Yaar! DY Aya! nae! aan a iF81, Mathematica Background Now, since Bly = —Byi one bas fy = 32. nba where eye isthe Levi-Cvita tensor in R?, and hence 1 Le B= 5D entide! nas! = pds? nds? + pds nds! + fads! nds? aay Wether ind a da = Wi dx" Adz? + fodx? Adx! + Wsdx! Adz? , (1.23) wh amt, a9 aay and hence the cartesian components of dar represent the curl of a vector field @ = (xs, 0,03): da <> y=curla “The exterior derivative off gives db => Peds! nds2ads? p= Bas! nat ad, (125) and hence, it represents the divergence ofthe vector field B = (Pi, a, ): dp <> divB ‘This way one recovers standard vector analysis in R?. Note that d(da) = 0, which reproduces the well-known identity diveuria Clearly, d(dp) = 0 since d(dp) as a four-form vanishes identically in R®, Moreover, if f is @ smooth function, then the identi curlgrad f = 0 follows from d(df) = 0 (see the previous Example). ° ‘These simple observations in R? may be immediately deduced from the following Proposition 1.1.2 The exterior derivative satisfies 1, Pa =d(da)=0, foranya e AM), 2. d@ AB) =da np +(-1ta rds, fora e AM). 1.1. Manifolds, forms and all that 9 Consider two manifolds M and N and let ¢ : M —> N bea smooth map. A pull-back operation induces a map ot MN) > AM) Proposition 1.1.3 The pull-back operation commutes with wedge product and exterior derivative, That is, Ga AB) = GAA, (1.26) and (da) =d(y"a), a2n for any differential forms « and fi on N. In the last equation we use the same letter ‘d’ 10 denote the exterior derivative on N and M. 1. 3 Integration of forms Differential forms occur implicitly in all branches of physics because they are natural objects appearing as integrands of line, surface, and volume integrals as well as their ‘n-dimensional generalizations. Consider R” with cartesian coordinates (x!,... x"). Having a function J : ” —> i, one defines an n-dimensional “volume” integral p> [save ff so! Clearly, the value ofthe integral can not depend upon the particular coordinates chosen to parametrize R°. In. particular, changing coordinates from (s',...,2") to G7") one finds f fa¥, with wx ds! da aV=sav, where J stands for the Jacobian of the transformation, i.e. o(%) What is the origin of /? Note that on R" we have a natural n-form dx! A... A dx" One immediately sees that J ABN. Nd = Idx! A. Adx", (1.28) which shows that the convenient notation dV = dx'...dx" actually denotes an n- form, and hence, it should rather be written as follows dV ads! a... Adx" ‘Hence, itis clear that to perform n-dimensional integration one needs n-forms,101, Mathematical Background Definition 1.1.4 An n-dimensional manifold M is orientable iff there exists a nowhere. vanishing n-form t on it One shows that a manifold M is orientable if we can cover it by coordinate patches (Wi.44) having positive Jacobians in each overlap ic. det(gji(x)) > 0 for any x € U; AUy, Itshould be clear that if M is orientable, then there are exactly two different ‘ways to orient it, OF course, if M can be covered by a single chart then itis orientable. For example, any open subset of Ris an n-dimensional orientable manifold Example 1.1.6 (Riemannian manifold) A Riemannian manifold (M, g) is a smooth, manifold M together with a smooth tensor g of type (0, 2), called a metrie tensor, such that 1. gissymmetric, 2, foreach x € M, the bilinear form gs : TyM x T;M —+ R is nondegenerate. ‘A Riemannian manifold is called proper if ax(v.0) > 0 forallve eM, v #0 Otherwise a manifold is called pseudo-Riemannian. For example, Euclidean space is proper Riemannian, whereas the Minkowski space > is pseudo-Riemannian, Note that on a Riemannian orientable manifold one may define a canonical volume form. In the space of differential forms on M one introduces so called Hodge operation (or Hodge star), + akan — artan. (129) by the following formula: Cine = VET eine eine ® (130) with g = det(g)) and ad seg gM mam asp Here ¢;,.j, stands for the Levi-Civita tensor in R", and g!? denotes the inverse of gi The form 3a is usually called the Hodge dual of a. A Hodge star induces a natural volume form 1.32) whee‘ cinta Ms, Hs) = ry Bey v= viglds! A... Ads". (1.33) 1 compat en wan = f+ se : is called the volume of M (with respect to r) ° LA. Manifolds forms and al that 11 Consider now an n-dimensional manifold M and let K be a k-dimensional orientable submanifold of M.! Denote by j : K <> M a canonical embedding. If a is @ k-form fon M, then ja is a k-form on K, and hence one may define an integral of "a over &: to — [ire Let (y!,... ,y®) denote local coordinates on K, and let the embedding j be described by x= alot, = 2G, io wom Oy) Then if a familiar form: fre Effet Bhat nondt, 035 which is a generalization of the line and surface integrals in R°. ‘To formulate one of the most important results in the theory of integration of differential forms we need a notion of a manifold with boundary. Let R". = {(x!,... x") eR" |x! > 0}. Then the boundary, aR, of Ris defined by JR" = ((x!,..., x") | x! = 0}. An n-dimensional manifold, M, has the structure of a manifold with a boundary when there exists an open covering (Uj, 91) such that g, (Ui) defines an open suibset of R's. The boundary, 8, of M is defined by am = Ue; (One sees that the boundary, 3M, of M is an (n ~ 1)-dimensional differential manifold Example 1.1.7 Let 8" be a unit ball in R" iyatedc" A... Adex!*, the above integral may be rewritten in a more ‘arty arafol er"! Pah? st} UA © is aLdimensionl submanifold if for any point x © K there exists a char (U, ) 09 M, sich vat elune UO K 88 0,....0)121, Mathematical Background ‘Then the boundary of B" is an (n ~ 1)-dimensional sphere, apt =st!, which agrees with our intuition of a boundary. ° Note thatthe notion of a boundary satisfies aM = aM) = (0), 036) for any manifold M, that is, a boundary does not have a boundary. In particular, the boundary of an n-dimensional sphere 4S" = 42"! = (@}. A manifold without a boundary is called closed. ‘Theorem 1.1.4 (Stokes theorem) Le M be an n-dimensional manifold with bound- ary, and let » € A"-!(M). Then where 0M denotes the boundary of M. 8 The Stokes theorem generalizes well-known theorems from vector 1, IF isa two-dimensional surface in R° and A.a vector field then flomacas= faa ‘where dS denotes a surface element on ¥:, and dl stands fora line element along the closed curve C = aE. 2. IV isa three-dimensional region in R? and A a vector field, then [waar =f aces. where dV denotes the volume element in R°. This formula is usually called a cs een 3 1.1.4 De Rham cohomology Recall from the vector analysis in R¥ that div (curl A) = 0 for any (smooth enough) vector field A. However, the converse statement isin general not tru, that i, the van- ishing of div B does not imply the existence of A such that B = curl A. In the case of AP EAM), a2 = a8. Clearly, ay ~ an => day =dap Evidently it is an equivalence relation, and hence we may define the space of equiva- lence classes HEM) = Z#(M)/BEM) aap